The hottest Risk modeling Substack posts right now

Estimated annual catastrophe losses rose sharply in nominal dollars from 2020 to 2025, but much of that increase is driven by greater exposure — more construction and higher construction prices — rather than clear evidence of climate-driven loss growth.
When losses are measured as a share of global GDP, weather-related disaster losses have not increased since 1990 and were below the long-term average in 2025.
Different firms produce different loss totals and insured-loss numbers are more reliable than total loss estimates; to detect climate signals you should look at climate data rather than economic loss data, and keeping disaster impacts low will require continued effort.

Property/casualty insurers are earning record profits and strong underwriting gains, so the industry is financially healthy despite media claims of collapse.
Underwriting returns swing up and down year-to-year but show no long-term trend, meaning insurers are not in a systemic underwriting crisis.
A booming climate‑risk vendor industry produces wildly different risk models, and those uncertain assessments have helped justify big insurance rate hikes even though the direct climate-driven increase in losses appears small.

The financial sector framed a new category called "climate risk" and built a regulatory and commercial ecosystem around it, treating it as a novel systemic threat to global finance.
That risk has been measured mainly by economic losses from extreme-weather events, which often mixes up rising damages with actual changes in weather rather than accounting for exposure and vulnerability.
Financial actors argue historical climate data is a poor guide and have pushed new scenarios, models, and private vendors to quantify "climate risk," creating a large market influence despite questions about the scientific basis.

Insurance companies are making record profits even as headlines claim a climate-driven insurance crisis, and recent premium hikes seem driven in part by rules to account for “climate risk” and the growth of risk-modeling services.
The issue presents data across many areas — dark oil tankers, moderates’ confidence in science, an energy skills gap, red-state/blue-state electricity price differences, southern-hemisphere wheat, and a comeback of climate-realist views.
More analysis is coming, including a follow-up on insurance and climate and a ranked list of major climate-research scandals, and the full material is available to paying subscribers.

Saying "no detectable harm" is the same as "zero harm" is misleading: biological repair can make extra radiation damage undetectable but not literally zero, so a tiny nonzero risk can remain.
Competing harm models (like LNT versus repair-aware models) produce vastly different low-dose risk estimates, so claiming absolute zero harm invites logical and rhetorical attacks and weakens your position.
For regulation and to avoid crippling tort liability, the industry needs a clear, quantitative radiation-harm model to calculate compensation and make nuclear power economically viable even when releases cause no detectable health effects.

Get a weekly roundup of the best Substack posts, by hacker news affinity: